Skip to content
finalstate.aiThe Frequency-Impact Matrix: A Triage Grid Dressed as a Ruler
102Atlas

102THE GRID ON EVERY WALL

The Frequency-Impact Matrix: A Triage Grid Dressed as a Ruler

A frequency-impact matrix sorts risks into categories. Its bins can hide different loss profiles; a consequential decision still needs assumptions, tested controls and an accountable owner.

A marked matrix cell connects to an unsigned review record.The selected category opens a review with an unresolved owner. No cell is labelled acceptable and no numeric measurement is inferred.ConsequenceFrequencyReviewOwner?
Conceptual triage grid. Cell position is an ordinal category, with no implied probability, loss amount or approval.

FOUR CORNERS

Four corners suggest different questions

A two-by-two matrix gives conditional questions for rare or frequent events and minor or severe consequences.Each cell contains a question: escalate, reduce, check appetite or standardise. The diagram does not mark three corners solved or reserve ownership for the rare severe corner.RareFrequentSevereMinorEscalate?Reduce?Withinappetite?Standardise?
Conditional prompts, not calibrated thresholds or automatic responses. Every corner can require an owner, evidence and applicable controls.

The rare severe corner includes exposures worth investigating. That does not make value exclusive to this corner: measured gross gains still need a retained-value account.

  • Frequent and minor: consider standardisation or automation
  • Frequent and severe: consider reduction, transfer or insurance
  • Rare and minor: check appetite and cumulative exposure
  • Rare and severe: examine scenarios, controls and ownership

BORN TO ESCALATE

A documented lineage, not an invention date

The original five-row, four-column example shows index 1D connected to a managing-activity decision.All twenty alphanumeric cells reproduce the first example's category codes. The highlighted remote/catastrophic cell requires a managing-activity decision under that example's suggested criteria; it is not a probability estimate or a universal acceptance rule.1984SeverityIIIIIIIVA1A2A3A4AB1B2B3B4BC1C2C3C4CD1D2D3D4DE1E2E3E4EFrequencyMA decisionManaging activity
MIL-STD-882B (1984), Appendix A, pp. A-3–A-5, fig. 1. The first example uses A–E for frequent, probable, occasional, remote and improbable; I–IV for catastrophic, critical, marginal and negligible. Index 1D is in the suggested “undesirable” group requiring a managing-activity decision. Historical example, not current guidance.

The original standard, Appendix A, uses examples to prioritise corrective action. Its managing activity retains specified review and decision responsibilities. This is one historical lineage of triage and escalation.

  • MIL-STD-882B, 30 March 1984
  • Example matrices: five frequency rows, four severity columns
  • Priority and managing-activity review remain distinct

Examine the next measure before justifying a stop

HSE’s Reducing risks, protecting people (2001), Appendix 3, explains the gross-disproportion test through Edwards v. National Coal Board (1949). This scoped account does not make ALARP a universal decision rule for every latent exposure.

AS LOW AS REASONABLY PRACTICABLE

A band does not demonstrate ALARP

Three conceptual tolerability regions sit beside an unassigned assessment of further controls.The assessment record opens without a risk marker descending to an approved stopping point. Its status stays unresolved; broadly acceptable does not mean that reasonably practicable duties disappear.UnacceptableTolerableonly if ALARPBroadlyacceptableFurthercontrols?Assessment unresolved
HSE, Reducing risks, protecting people (2001), paras. 122–129, printed pp. 42–45; Appendix 3. Conceptual regions, not numeric bands. No case is assigned to a region or certified as ALARP.

HSE’s current regulatory explanation retains the tolerability framework for its Great Britain workplace health and safety role. It expressly says the framework is not dutyholder guidance. The decision still depends on the applicable duties and evidence.

  • Unacceptable: activity generally ruled out unless modified or exceptional reasons apply
  • Tolerable: assess, reduce as low as reasonably practicable, review
  • Broadly acceptable: reasonably practicable duties still apply

UNFOLD THE CELL

Inspect the paths and the barriers

Preventive barriers interrupt paths before the top event, while recovery barriers sit after it; one preventive bypass opens and one recovery owner remains unresolved.The path through the bypass becomes visible without quantifying failure likelihood. The recovery owner question is a separate missing governance record, not the cause assigned to that bypass.ThreatsConsequencesTopeventPreventRecoverBypassOwner?
Conceptual bowtie, following CAA’s method. The bypass and unanswered owner field are illustrative and distinct: a missing owner is not a measured failure probability. Bar positions imply no verified effectiveness.

The CAA bowtie method separates threats, the top event and consequences. An escalation factor can weaken a control. For a latent exposure, drawing the barrier does not establish that it works.

  • Threats can lead to a loss-of-control top event
  • Prevention acts before the top event
  • Recovery limits consequences after it
  • Record weaknesses, ownership and test evidence

ONE LINE INSTEAD

Build the loss profile from stated inputs

A fixed step curve from invented inputs shows a 0.1 percent annual chance above two million euros and 0.01 percent above twenty million, with appetite unspecified.The complete curve is derived from the three stated hypothetical annual outcomes. Highlights inspect existing points without changing values. Vertical segments mark jumps; they are not intermediate probabilities. The financial profile does not establish counterfactual prevention or ALARP.Annual P(L > x), %00.010.050.10310202530Loss threshold x, €m€2m: 0.1%€20m: 0.01%Appetite unspecified
Fictional annual distribution, shared with node 009. Linear axes show threshold x in €m and P(L > x) in percent. The curve drops at €3m and €25m; upper endpoints there are open and lower endpoints closed. No appetite boundary, model validation or legal compliance is implied.

The inputs come from the illustrative tank case, not from the selected cell. The curve makes assumptions visible beyond the bins. It still cannot prove a particular avoided loss.

  • Fictional annual losses: €0 at 99.9%, €3m at 0.09%, €25m at 0.01%
  • P(L > €2m) = 0.1%; P(L > €20m) = 0.01%
  • Appetite remains unstated; no acceptance follows

BOTH WAYS AT ONCE

An opportunity is a candidate for a decision

Threat and opportunity grids lead to separate decision records; an opportunity candidate has an unanswered rights field.The opportunity marker is a candidate only. Its record opens for inspection of rights without becoming a secured option or measured gain. Neither wing is automatically accepted or assumed symmetric.ThreatOpportunityLikelihoodHarmBenefitResponse?Rights?
Conceptual comparison, informed by Hillson (2002), abstract. Cell dimensions and matching positions imply no equal likelihood, value or response. Rights remain to be established.
  • Hillson proposed one process for threats and opportunities
  • Downside and upside need their own evidence
  • A candidate is not a secured preferential right

Hillson’s 2002 paper00074-6) proposes extending the risk process to opportunities. This mirrored comparison is explanatory. Identifying potential upside does not secure the preferential rights of a real option.

The cell is the question, not the answer.

Cox’s 2008 analysis identifies tied ratings and possible ranking errors. The specific constructions and their limits belong with range compression; they are not a universal failure rate for every matrix.

THE APPARATUS THAT DECIDES

Attach the evidence to the decision

A selected cell connects to a decision file with owner, test and rule questions; the decision remains pending.The file opens as the evidence requirements are named. Their blank fields remain unresolved rather than being silently ticked off. Pending is a status of this illustrative review, not a rule that every risk needs every instrument.DecisionRule?Owner?Test?Pending
Illustrative decision file. Missing owner, test evidence and applicable rule remain visible; no automatic approval is issued.

Controls and evidence can change the cost of responding. Carry their assumptions into the decision about the marked cell, including the next review and the person authorised to reopen it.

  • State the rule and its scope
  • Check barrier ownership and relevant tests
  • Record model assumptions and the authorised decision
Read the transcript

01 · THE GRID ON EVERY WALL

A frequency-impact matrix sorts risks into categories. One axis describes how often an event might occur; the other describes its consequences. Before placing a mark, define the event, the time horizon and the category boundaries. Then the grid can direct attention or trigger a review. The mark does not contain a full loss profile, demonstrate a working control, or appoint the person who can decide. Follow one selected cell to its review record. The first missing field is the owner. The review remains open.

02 · FOUR CORNERS

The four corners are a sorting vocabulary. Frequent and minor losses may be candidates for standardisation or automation. Frequent and severe losses may call for reduction, transfer or insurance, subject to what those measures cover. Rare and minor losses still need to fit appetite, including their accumulation with other exposures. Rare and severe events deserve examination of scenarios, controls and who can act. A frequent risk also needs ownership; a rare event may still be rehearsable. The grid does not tell you that any corner is solved.

03 · BORN TO ESCALATE

One documented lineage is military. The nineteen eighty-four revision B of United States military standard eight eighty-two includes example hazard matrices with five frequency rows and four severity columns. Their codes help prioritise corrective action. In the first example, index one D combines remote frequency with catastrophic severity. Its suggested criteria require a managing-activity decision. The managing activity is the acquisition organisation, or a contractor imposing safety tasks on suppliers. The standard explicitly warns that its examples do not fit every programme. It documents one use, not the invention of the instrument.

04 · THE COMPUTATION

HSE's account of reasonable practicability asks about a particular measure. In Reducing risks, protecting people, it explains the nineteen forty-nine Edwards case by comparing the risk reduction with the sacrifice involved, including money, time and trouble. A duty-holder seeking to rule out a measure must establish gross disproportion, with a bias toward safety. This is not a fixed multiplier or an ordinary break-even rule. The legal context matters. A firm's financial appetite cannot substitute for its applicable health and safety duties. The comparison asks for evidence about the next control and a reasoned basis for stopping.

05 · AS LOW AS REASONABLY PRACTICABLE

HSE's tolerability framework has three conceptual regions. An unacceptable risk generally rules out the activity unless it can be modified or exceptional reasons apply. In the tolerable region, the risk must be assessed, controlled as low as reasonably practicable and periodically reviewed. Broadly acceptable risks do not erase the duty to reduce risk where reasonably practicable or otherwise required by law. Open the assessment of further controls before locating a case in this framework. A coloured band supplies none of that evidence. In this exhibit, the assessment remains unresolved. The framework informs HSE's regulatory decisions; its current explanation expressly says it is not dutyholder guidance.

06 · UNFOLD THE CELL

A bowtie exposes the paths a cell compresses. Threats lead toward the top event, the moment control over a hazard is lost. Consequences follow it. Add preventive barriers on the left: they act before the top event. Add recovery barriers on the right: they can reduce the chance or severity of a consequence after control is lost. Now inspect what can weaken them. In the illustration, a bypass opens through one preventive barrier. A separate recovery barrier has no owner recorded. Those are different gaps. Check the actual control, its responsible owner and the evidence from a relevant test; none is established by drawing the bar.

07 · ONE LINE INSTEAD

A loss-exceedance curve requires inputs beyond the matrix. Use the fictional annual tank distribution from the adjoining case: no loss with ninety-nine point nine percent probability, a three-million-euro loss with zero point zero nine percent, and a twenty-five-million-euro loss with zero point zero one percent. The horizontal axis is a loss threshold; the vertical axis is the annual chance of exceeding it. Above two million euros, the chance is zero point one percent. Above twenty million, it is zero point zero one percent. The step drops at three million and twenty-five million. Appetite is still unspecified. The curve exposes this assumed profile; it validates neither the inputs nor an avoided loss, and it establishes no legal compliance.

08 · Advertisement · ElevenLabs Studio

The last mile of content is often sound: the right voice, the clean take, the version that works in another market. ElevenLabs brings AI voice, music, dubbing, and sound tools into one production flow, so teams can shape audio with the same care they bring to the story. Create the version people can actually hear.

09 · BOTH WAYS AT ONCE

The process can also examine opportunities. David Hillson's two-thousand-and-two paper proposed extending risk management to include uncertainties with beneficial effects on objectives. A second wing can make those candidates visible beside threats. The symmetry is a drawing choice, not a claim that the two sides have equal probabilities, values or responses. Select an opportunity and inspect its rights. Does the organisation have an enforceable route to act, or has it merely noticed something attractive? Potential upside alone is neither retained value nor a secured preferential right.

10 · THE QUESTION, NOT THE ANSWER

A cell can direct a review while concealing differences that matter to the decision. Cox's analysis of risk matrices identifies tied ratings for quantitatively different risks and the possibility of reversed rankings. Those results depend on the construction being examined. They do not mean that every matrix fails in the same way. Keep the category definitions and judgments visible, then inspect the evidence needed for the next action. The cell is the question, not the answer.

11 · THE APPARATUS THAT DECIDES

Attach the evidence to the decision. For the marked cell, identify the applicable rule, the available action and the person authorised to choose it. Check who owns the relevant barrier and what its test actually examined. If a loss model informs the choice, retain its assumptions and uncertainty. A scenario can expose another path. Not every risk needs every instrument, but the chosen method must answer the decision at hand. Here, owner, test evidence and rule remain unresolved, so the file stays pending. Record what would reopen the review, then carry the case forward.

01 / 11 · THE GRID ON EVERY WALL0:00 / 7:40